Re: [squid-users] Question about cache_peer, squid 3.2 Bug ?

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Thu, 13 Sep 2012 11:44:26 +1200

On 13.09.2012 05:31, FredB wrote:
> Hello,
>
> I'm trying to migrate squid 3.0 STABLE 25 to Squid 3.2.0.19, all
> seems works except one very strange thing.
>
> There is a cache_peer like this
>
> cache_peer 10.x.x.x parent 3128 0 no-query login=PASS proxy-only
> no-digest no-netdb-exchange - (10.x.x.x run with basic ident)
> After this there are another rules and cache_peers who work without
> problem - But without any ident rules ...
>
> At the end, my last rule is (default)
> cache_peer_access 10.x.x.x allow all
>
> The problem is when some users enter in this last rule, there is no
> packet in 10.x.x.x and squid log this
> [12/Sep/2012:16:54:04 +0200] "GET
> http://www.google.com/m/gne/suggest/v2? HTTP/1.1" 502 1213
> TCP_MISS:ANY_OLD_PARENT "Android/1.0 (GT-I9100P IML74K)"
>
> No problem with another user, for example
>
> [12/Sep/2012:17:31:15 +0200] "GET
> http://www.wikipedia.fr/Fichiers/appel_au_don.gif HTTP/1.1" 304 414
> TCP_MISS:FIRSTUP_PARENT "HTC_TyTN_II Mozilla/4.0 (compatible; MSIE
> 6.0; Windows CE; IEMobile 7.11)"
>
> If I switch to 3.0 the problem disappears, it's hard to debug because
> I can't reproduce with my phone (no problem for me)

Please update to the stable release 3.2.1.

With 3.2 you have "debug_options 11,2" which will produce a full trace
of HTTP headers in cache.log which will greatly help understanding why
the 502 occured. (careful about size and confidential info if this is
used on a production machine)

I also suggest using the squid format for access.log, it includes
details such as the IP of the server which that 502 is related to.

Amos
Received on Wed Sep 12 2012 - 23:44:29 MDT

This archive was generated by hypermail 2.2.0 : Mon Sep 17 2012 - 12:00:03 MDT